tiagozip/truthsocial

Truth Social scrape

hundreds of millions of public posts from over 500k Truth Social accounts since February 2022, scraped continuously since early 2026 and updated monthly.

from datasets import load_dataset

posts = load_dataset("tiagozip/truthsocial", "posts", split="train", streaming=True)
accounts = load_dataset("tiagozip/truthsocial", "accounts", split="train")

format

config	rows	notes
posts	100M+	one file per month, data/posts/YYYY-MM.parquet
accounts	500k+	profile snapshot, one row per account
groups	8k+	Truth Social's group feature

posts

column	type	notes
id	uint64	snowflake, (unix_ms << 16) | hash16
created_at	timestamp[ms, UTC]
account_id	uint64	joins to accounts.id
acct	string	handle at fetch time
display_name	string
content	string	HTML format
language	string	as declared by the client, often wrong
in_reply_to_id	uint64	0 when not a reply
quote_id	uint64	0 when not a quote
reblog_of_id	uint64	0 when not a retruth
group_id	uint64	0 when not posted to a group
group_name	string
tags	list[string]
mentions	list[string]	handles
media_urls	list[string]	these may be deleted
media_types	list[string]
replies_count	uint32	counter at fetch time, not current
reblogs_count	uint32	same
favourites_count	uint32	same
upvotes_count	uint32	same
visibility	string
sensitive	uint8
source	string	which timeline this post was fetched from
fetched_at	timestamp

accounts

id, acct, display_name, note (bio HTML), created_at, followers_count, following_count, statuses_count, verified, bot, location, website, avatar, last_status_at, fetched_at.

what this is not

• not complete. Truth Social has roughly 1.1 billion reachable posts, so consider this a large recent sample. the most recent months are the most complete.

• engagement counts are snapshots. every counter is whatever it read at fetched_at. a post fetched minutes after it was stored will likely show near-zero engagement. do not treat these as final values.

• followers and following are absent. due to scraping limitations, social graphs are not enumerable.

license

tiago.zip/license, version 1.2. in short:

• research, publishing, and training or evaluating ML models are all fine, and none of it needs approval.
• commercial use is fine too, but email hi@tiago.zip first saying who you are, which dataset you're using and what for.
• credit "Data from tiago.zip/datasets" wherever people see your work, whether that's a paper, a post, an app or a model card.
• don't republish, sell, mirror or rehost the data. link to tiago.zip/datasets instead.
• don't use it to identify, track, profile or harass anyone.
• when you get a removal notice, delete those records from your copies within seven days.

disclaimers

access can be revoked at any time and all data is provided as-is. this dataset contains personal information about real people, a lot of whom are not public figures. this dataset is not a reflection of the views of the author or the institutions they are affiliated with. please be respectful. every post in this dataset was publicly readable at the time it was fetched, and all scraping was done unauthenticated.

TRUTH Social and associated marks are trademarks of their respective owners. this dataset is independent and unaffiliated.

opting out

you may opt out at optout.tiago.zip or by adding "#noindex" to your truth social bio.

citation

@misc{truthsocial,
  title  = {truthsocial: a corpus of public Truth Social posts},
  author = {tiagozip},
  year   = {2026},
  url    = {https://huggingface.co/datasets/tiagozip/truthsocial}
}